Learning deep representation for trajectory clustering

Abstract

Trajectory clustering, which aims at discovering groups of similar trajectories, has long been considered as a corner stone task for revealing movement patterns as well as facilitating higher level applications such as location prediction and activity recognition. Although a plethora of trajectory clustering techniques have been proposed, they often rely on spatio‐temporal similarity measures that are not space and time invariant. As a result, they cannot detect trajectory clusters where the within‐cluster similarity occurs in different regions and time periods. In this paper, we revisit the trajectory clustering problem by learning quality low‐dimensional representations of the trajectories. We first use a sliding window to extract a set of moving behaviour features that capture space‐ and time‐invariant characteristics of the trajectories. With the feature extraction module, we transform each trajectory into a feature sequence to describe object movements and further employ a sequence‐to‐sequence auto‐encoder to learn fixed‐length deep representations. The learnt representations robustly encode the movement characteristics of the objects and thus lead to space‐ and time‐invariant clusters. We evaluate the proposed method on both synthetic and real data and observe significant performance improvements over existing methods.

What makes trading strategies based on chart pattern recognition profitable?

“…The DTW algorithmic technique is based on dynamic programming and aims to find the optimal alignment between two sequences (that may differ in length) by dynamically stretching and compressing their time axis. Although it initially became popular in speech recognition applications (Gomez‐Donoso, Cazorla, Garcia‐Garcia, & Garcia‐Rodriguez, 2016; Sakoe & Chiba, 1978), it was later applied to other research areas (Yao et al, 2018), including finance (e.g., Wang, Xie, Han, & Sun, 2012).…”

Vehicle trajectory modelling with consideration of distant neighbouring dependencies for destination prediction

“…Wu et al (2017) treat a trajectory as a sequence of transition state from one edge to another and transform each state into an embedding vector to get dense representations of sequence units. Yao et al (2018) extract a set of moving behaviour features with a fixedlength sliding window and then learn a fixed-length representation of a trajectory through a sequence-to-sequence model. Although these methods can model different movement characteristics in trajectories, it remains a challenge to transform trajectories into lowdimensional representations simultaneously retaining dependencies between visited locations that can cover a long distance or a short distance.…”

Clustering of Driving Encounter Scenarios Using Connected Vehicle Trajectories

“…The authors in [4] also developed a two-step procedure to predict a trip's destination using a density-based clustering of destination and the initial part of trajectories. Yao et al [12] used a sliding window to extract a set of moving behavior features that capture space and time-invariant characteristics of the trajectories. Zhao et al [13], [14] extracted lane change behavior according to the vehicle's raw movement trajectory and conducted the scene-based analysis.…”